What can your hardware do?
The model, the precision, the workload and the container are fixed. Only your hardware changes. The numbers compare.
For hardware that you can run at home. LocalMax may accept a result from a data-centre GPU or system, but it reserves the right not to show it and not to rank it.
Read the code first
Source on GitHubLocalMax runs on your machine. Examine the source before you run it. Every part is public: the runner, the container files, the profiles and the checks.
GPU model, VRAM, driver, CUDA, PCIe link, power limit and clocks. CPU model and core count. RAM. OS and kernel. All measured values, the raw records for each request, and the telemetry.
Hostname. Username. File paths. GPU serial numbers. Board UUIDs. MAC addresses. Environment variables. Data about other processes. The runner removes secrets and paths from logs on your machine, before it writes them.
LocalMax gives your machine a name such as Rocinante-K7X2P. The name comes from a key on your machine. You do not type a name. Keep the 5-character code. Use it to find your results at /systems/CODE.
Submission is voluntary. You control what you send. Commandlocalmax inspect shows the full content before upload. Legal · Privacy
Three tiers
One model cannot span 12 GB to 192 GB. A tier is the VRAM its run must fill. Each tier has its own model.
A tier gives the largest model it runs, in billions of parameters, at FP8. LocalMax also accepts and ranks INT4 and NVFP4 submissions. Each lane has its own leaderboard, because different precisions are different work.Methodology →